T1ρ mapping improvement using stretched-type adiabatic locking pulses for assessment of human liver function at 3T

(1)

T1ρ mapping improvement using stretched‑type adiabatic locking pulses for assessment of human liver function at 3T

著者奥秋知幸

著者別表示 Okuaki Tomoyuki journal or

publication title

博士論文本文Full 学位授与番号 13301甲第4616号

学位名博士（保健学）

学位授与年月日 2017‑09‑26

URL http://hdl.handle.net/2297/00050302

doi: 10.1016/j.mri.2017.03.006

Creative Commons : 表示 ‑ 非営利 ‑ 改変禁止 http://creativecommons.org/licenses/by‑nc‑nd/3.0/deed.ja

(2)

T

₁_ρ

mapping improvement using stretched-type adiabatic locking pulses for assessment of human liver function at 3 T

Tomoyuki Okuaki

^a,b,

⁎ , Yukihisa Takayama

^c

, Akihiro Nishie

^d

, Tetsuo Ogino

^e

, Makoto Obara

^e

, Hiroshi Honda

^d

, Tosiaki Miyati

^b

, Marc Van Cauteren

^a

aPhilips Healthcare, Tokyo, Japan

bDivision of Health Science, Graduate School of Medical Sciences, Kanazawa University, Kanazawa, Japan

cDepartment of Radiology Informatics and Network, Kyushu University, Graduate School of Medical Sciences, Fukuoka, Japan

dDepartment of Clinical Radiology, Kyushu University, Graduate School of Medical Sciences, Fukuoka, Japan

ePhilips Electronics Japan, Tokyo, Japan

a b s t r a c t a r t i c l e i n f o

Article history:

Received 30 November 2016

Received in revised form 17 February 2017 Accepted 25 March 2017

Available online xxxx

Purpose:The purpose of this study is to investigate the performance of stretched-type adiabatic spin lock pulses for homogeneous spin locking with aﬂexible spin lock time (TSL) setting.

Methods:T1ρvalues were obtained from 61 patients andfive normal volunteers who were categorized using the Child–Pugh classification and scanned using each spin lock pulse type. The pulses used were the block and two kinds of hyperbolic secant (HS); HS8_10, and HS8_5. Visual scoring was categorized using a four point scale (1:Severe, 2:Moderate, 3:Mild and 4:None) to evaluate the homogeneity of the T_1ρmap and the source images obtained by each spin lock pulse. Mean T_1ρvalues among the patient groups with different Child–Pugh classification were compared.

Results:The visual assessment scores were 1.98 ± 1.05 for block pulse locking, 3.87 ± 0.39 for HS8_10 pulse locking, and 3.83 ± 0.45 for HS8_5 pulse locking, respectively. The scores between block pulse and HS8_10 were signiﬁcantly different (pb0.001), as were those between block pulse and HS8_5 (pb0.001).

The median T_1ρvalues of normal liver function, Child–Pugh A, and Child–Pugh B or C were 37.00 ms, 40.77 ms, and 42.20 ms for block pulse, 46.75 ms, 50.78 ms, and 55.60 ms for HS8_10, and 48.80 ms, 55.42 ms, and 57.80 ms for HS8_5, respectively.

Conclusion:The spin locking sequence using stretched-type adiabatic pulses provides homogeneous liver T_1ρ maps with reduced artifact and is necessary for a robust evaluation of liver function using T1ρ.

Keywords:

T1ρ

stretch type adiabatic pulse liver

spin lock hyperbolic secant

1. Introduction

Assessment of liver function is essential for the management of patients with liver disease and for the prevention of postoperative hepatic failure. Liver function is frequently estimated by measuring biochemical parameters in the blood, such as bilirubin, aminotransferase, alkaline phosphatase, gamma-glutamyl transferase, albumin, and prothrombin activity[1]. The Child-Pugh classiﬁcation[2,3]is based on a combination of serum albumin, serum bilirubin, prothrombin activity, ascites, and hepatic encephalopathy to reﬂect liver functional reserve more accu- rately than by using any of these biological factors alone[3]. Previous

reports using magnetic resonance imaging (MRI) state that T1 mapping and T2* mapping can be used to assess liver function[4,5]. Liverﬁbrosis is a common feature of most chronic liver diseases and ultimately pro- gresses to liver cirrhosis with the accumulation of proteoglycans and collagen and other macromolecules in the extracellular matrix[6–8].

Conventional MRI cannot evaluate liverfibrosis directly. Liver biopsy is carried out for the diagnosis and monitoring of progression as a standard of reference; however, liver biopsy is an invasive procedure, is prone to error, and involves the risk of complications[9,10]. Another ap- proach, magnetic resonance elastography (MRE), is a noninvasive method for the detection offibrotic liver and staging of liverfibrosis [9–11]; however, additional equipment is required for MRE, especially the transducer to generate a vibrational wave[12].

Recently, T_1ρmeasurement has been widely applied to investigate diseases of the cartilage[13–17], prostate, disc[18], myocardium[19], and liver[7,8,20–23]. With regard to the liver, Wang et al. reported that in an animal model T_1ρrelaxation was used successfully to detect early liverﬁbrosis and that increased liver collagen results in an increase Abbreviations:MRE, magnetic resonance elastography; RF, radio frequency; TSL, spin

lock time; HS, hyperbolic secant; MLEV, Malcolm H. Leviti; IDL, Interactive Data Language; ROI, region of interest.

⁎ Corresponding author at: Philips Healthcare , 13-37, Kohnan 2-chome, Minato-ku, Tokyo, 108-8507 Japan.

E-mail address:tomoyuki.okuaki@philips.com(T. Okuaki).

Contents lists available atScienceDirect

Magnetic Resonance Imaging

j o u r n a l h o m e p a g e :w w w . m r i j o u r n a l . c o m

(3)

of the T_1ρvalue[20]. Contrary to that report, Sirlin et al. suggested that T_1ρrelaxation does not directly reﬂect liverﬁbrosis[8]. Moreover, Allkemper et al. and Rauscher et al. reported that T_1ρvalues were prolonged proportional to the progression of the Child–Pugh grade and liver cirrhosis[24,25].

In most of these studies, a block pulse was used as a spin lock pulse.

However, at 3 T, severe artifacts due to B0 and B1 inhomogeneity were observed, especially for a larger organ like the liver. Furthermore, these artifacts not only cause artefactual T1ρimages but also hamper accurate measurement of the T_1ρ. Such artifacts are caused by inhomogeneous B0 and B1 and therefore imperfect excitation pulses in the spin lock pulse train. Witschey et al.[26]reported an improved spin locking pulse, namely the echo locking method, to remediate these artifacts in a phan- tom and the human brain but not in the liver.

Adiabatic pulses are characterized by the simultaneous modulation of radio frequency (RF) wave amplitude and frequency. Well-chosen modulation wave forms result in insensitivity to a broad range of B0 and B1 inhomogeneity[27–31].

A few papers reported application of adiabatic pulses for spin locking. Taheri S et al. and Michaeli S reported a simulation study[32, 33], Mangia S et al. studied the brain at 4 T[34], Casula V et al. studied knee cartilage[35]and Yang Q et al. applied them to liver[22].

The purpose of this study was to investigate the performance of the stretched-type adiabatic spin lock pulses for homogeneous spin locking with aﬂexible spin lock time (TSL) setting. Moreover, we aimed to investigate the clinical usefulness of the T_1ρvalue acquired using the improved spin locking pulses for the assessment of liver function.

2. Materials and methods 2.1. Subjects

This study was approved by our Institutional Review Board; it com- plied with the standards of the Ethics Committee.

Between June 2013 and July 2014, 69 patients underwent MRI of the liver, including T1ρmaps, because a liver tumor was suspected due to chronic liver disease or malignant disease in other organs. We excluded eight patients with difficulty to determine Child–Pugh classification. Of the remaining 61 patients, 51 patients were categorized as Child–Pugh A (age range, 40–83 years; mean age, 63.4 years) and 10 as Child– Pugh B or C (age range, 43–74 years; mean age, 61.3 years). As the control, five normal volunteers (age range, 50–55 years; mean age, 52.6 years) were scanned.

2.2. Imaging protocols

MRI data acquisition was performed on a 3 T clinical scanner (Achieva TX; Philips Healthcare, Best, Netherlands). A 32-channel phased-array receiver coil was used.

Three types of rotary echo spin lock pulses were generated to achieve T_1ρ-weighted images (Table 1). The non-selective block pulse was used as a conventional spin lock pulse (Fig. 1). The spin lock pulse amplitude was set at 500 Hz, following the literature[20,36]; the TSLs were 1, 20, and 40 ms. Two different types of spin lock pulses type

were used, namely the block pulse and the stretched-type adiabatic preparation pulses.

The hyperbolic secant (HS) pulse, known to be insensitive to variations of B1 intensity approaching several orders of magnitude, was used as a base. The time variation of the wave form modulations was modiﬁed to further optimize the performance of the pulse.

An effective magneticfield (ωeff=γBeff) is induced by the pulse, whereBeffis the effective magneticfield vector andγis the gyromagnet- ic ratio. The time-dependent magnitude of the effectivefield is:

ωeffð Þ ¼t ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ω²1ð Þ þt ðω0−ωRFð Þt Þ² q

where;

ω1(t) =ω1maxsech(βtⁿ), (ω0−ωRF(t))= A∫sech²(βτⁿ)dτ, whereω1max

is the maximum amplitude. A determines the frequency sweep amplitude and is equal toμβ.μis a dimensionless parameter andβis a modulation angular frequency.ω0is the on-resonance frequency equal toγB0andωRFis the frequency modulation. For the original HS pulse, n equal to 1, whereas the stretched adiabatic pulse is generated bynN1 (Fig. 2). We applied a high n factor (n= 8) to generate two types of adiabatic spin locking pulse from the original pulse, which was characterized byμ= 5 andβ= 4 rad/s. The parameters of theﬁrst pulse type were pulse duration 10 ms, frequency sweep 636.6 Hz, and maximum amplitude 6.37μT (HS8_10). The second pulse type had a pulse duration 5 ms, frequency sweep 1273.2 Hz, and maximum amplitude 13.48μT (HS8_5). The Beffchange during the two types of adiabatic pulse is indicated inFig. 3; the average Beffﬁeld of HS8_10 and HS8_5 was 500 Hz and 1000 Hz, respectively. For these adiabatic pulses, the rotating frame relaxation time was measured using a pulse module consisting of consecutive identical HS pulses.

The number of pulses was increased to obtain the T_1ρimages with different TSLs forﬁtting from two to four for HS8_10 and from four to eight for HS8_5, with MLEV phase cycling[37](Fig. 4). The TSL was set to 0, 20, and 40 ms by increasing the number of pulses. TSL = 0 represents the absence of a spin lock pulse.

Table 1

Design parameters for the spin lock pulses.

Spin lock pulse Pulse type

Pulse duration per pulse

(ms) Frequency sweep (Hz)

Spin lock frequency (Hz)

B1max (μT)

TSL (ms)

Block Block – – 500 11.7 1,20,40

HS8_10 Stretched-type adiabatic (n= 8) 10 636.6 500.0* 6.73 0,20,40

HS8_5 Stretched-type adiabatic (n= 8) 5 1273.2 1000.0* 13.48 0,20,40

*average effective SL frequency.

Fig. 1.Scheme of the T1ρspin lock RF cluster using block pulse. The RF pulse train is 90(+x)–SL(+y)–SL(−y)–90 (−x). Magnetization is nutated into thexaxis by the initial 90° pulse. The magnetization vector is then spin locked by two block pulses, lockingﬁrst along the +yand then the–yaxis. Finally, the magnetization is nutated back into thezaxis by the 90° pulse along the–xaxis.

(4)

We simulated the efﬁciency of these three types of spin lock pulse in function of B0 and B1 using Bloch equations at TSL 20 ms (Fig. 5). The simulation was based on the Bloch-equations, solved using the ordinary differential equation function of Matlab (Mathworks, Natick, MA, USA).

We did not take relaxation effects into account. Theﬁgures show the ﬁnal magnetization, in function of B0 and B1, after presenting each type of spin lock pulse combination to the equilibrium magnetization.

The spin locking block pulse combination nutates the equilibrium magnetization into the x axis by the initial 90° pulse and is then spin locked by two block pulses that have opposite phases. After the spin locking the magnetization is nutated back into the z axis by a−90°

pulse (Fig.5a).

For the two types of stretched-type adiabatic pulse, the magnetization is spin locked by an even number of adiabatic pulses (Fig, 5 b and c).

In case of perfect locking, and ignoring relaxation, theﬁnal magnetization is equal to the equilibrium magnetization.

In theﬁgures we show theﬁnal magnetization as a fraction of the equilibrium magnetization, in function of both B0 and B1.

For imaging, a three-dimensional turboﬁeld echo sequence using the parallel imaging technique was employed with breath holding.

Volume shimming was used to minimize B0 inhomogeneity. Other imaging parameters were as follows: repetition time 2.1 ms, echo time 0.98 ms,ﬁeld of view 360 × 306 mm², matrix 256 × 205, slice thickness 10 mm, slice gap 0 mm, number of slices 3, number of acquisitions 1, shot interval 5000 ms, and scan time of T_1ρ-prepared image with each TSL 11.7 s. Scanned slices were set at the level of the hepatic hilum.

2.3. T_1ρimage analysis

T_1ρmaps were generated on a pixel-by-pixel basis using in-house developed software program written in Interactive Data Language (IDL 6.3; ITT, Boulder, CO) using a mono-exponential decay model:

M(TSL) = M0 × exp.(−TSL/ T_1ρ), where M0 and M(TSL) denote the equilibrium magnetization and T_1ρspin lock prepared magnetization for each TSL, respectively. T_1ρmaps were generated for each spin lock type with a Levenberg–Marquardtﬁtting algorithm[38]. For quantiﬁca- tion of liver T_1ρvalue, three regions of interest (ROIs) were manually placed on the liver parenchyma; the sizes of the ROIs were as large as possible, approximately 50–150 mm², avoiding blood vessels, tumors, and artifacts.

2.4. Visual assessment

Evaluation of the homogeneity of the T_1ρmaps and the T_1ρsource images was scored through visual evaluation by two MR clinical scien- tists (T.O. and M.O.) with 10–16 years of experience. Visual scoring was categorized using a 4-point scale as 1 severe, 2 moderate, 3 mild, and 4 none. The deﬁnition of visual scoring was as follows:

1: Severe: Severe artifacts, difﬁcult to evaluate T_1ρvalues;

2: Moderate: Mildly severe artifacts. Artifacts should be avoided to measure the T1ρvalue, however, theﬁtting to calculate T1ρvalues is acceptable using a limited area;

3: Mild: Some artifacts on the image;

4: None: No artifacts on the image.

2.5. Statistical analysis

We employed the Tukey–Kramer method to compare visual scoring between the image qualities and the Friedman test to compare the mean T1ρvalues of each spin locking pulse (the block pulse, HS8_10, and HS8_5).

Mean T_1ρvalues among the patient groups with different Child– Pugh classiﬁcations were compared using the Kruskal–Wallis test followed by Tukey's post hoc test. Statistical analysis was performed Fig. 2.Scheme of the amplitude and frequency modulation function for HS1 and HS8. HS8

is the stretched-type version of HS1. Time is shown on the horizontal axis.

Fig. 3.The amplitude, frequency and effective frequency change during two types of adiabatic pulses: HS8_10 a) and HS8_5 b).

(5)

using EZR free software (Saitama Medical Center, Jichi Medical Univer- sity, Saitama, Japan)[39]. APvalueb0.01 was considered signiﬁcant.

3. Results

The simulation suggested that the HS8_5 pulses would perform better as a locking pulse, given that for a broad range of B0 and B1 values the equilibrium magnetization is restored after the locking.

Two cases of typical source images and T_1ρmaps are shown inFig. 6a and b. There were artifacts on most of the spin locking images using the block pulse; however, the artifact level on the source images and T_1ρ mapping obtained with the adiabatic spin locking pulses was signiﬁ- cantly reduced.

3.1. Visual assessment

The intraclass correlation coefﬁcient of visual scoring between the two readers was 0.90, indicating good concordance in the readers' visual scoring.

The visual assessment score (mean ± standard deviation) of the homogeneity of the T_1ρmaps resulted in 1.98 ± 1.05 for block pulse locking, 3.87 ± 0.39 for HS8_10 pulse locking, and 3.83 ± 0.45 for HS8_5 pulse locking, respectively. Both types of adiabatic spin locking- derived maps scored signiﬁcantly better than the block pulse locking- derived maps (Fig. 7). The visual assessment score showed signiﬁcant differences between the block pulse and HS8_10 (Pb 0.001) and

between block pulse locking and HS8_5 (Pb0.001). There was no significant difference between HS8_10 and HS8_5 (P= 0.792).

3.2. T_1ρvalues

It was not possible to draw an ROI avoiding the artifacts in all source images for some T_1ρmaps using block pulse because artifacts appeared in different locations in each source image. This resulted in a limited area where the T_1ρvalue could be calculated to generate a reliable T_1ρ map. In contrast, there were no severe artifacts in most of the source images using adiabatic pulses.

The T1ρvalues calculated using each spin lock type are shown in Table 2. The mean T_1ρvalues for normal tissue of block, HS_10, and HS_5 were 37.07 ms (range: 32.47–42.41 ms), 46.12 ms (range:

44.03–48.77 ms), and 49.99 ms (range: 47.97–51.43 ms), respectively.

T_1ρvalues were signiﬁcantly different between each spin lock type (Friedman test,Pb0.0001).

The T_1ρvalues for each spin lock type are shown inFig. 8. The median T_1ρvalues of normal liver function, Child–Pugh A, and Child–Pugh B or C were 37.00 ms, 40.77 ms, and 42.20 ms for block pulse; 46.75 ms, 50.78 ms, and 55.60 ms for HS8_10; and 48.80 ms, 55.42 ms, and 57.80 ms for HS8_5, respectively.

T_1ρ values were not significantly different between normal liver function and Child–Pugh B or C using block pulse (Kruskal–Wallis, P= 0.038). We found a significant difference between normal liver function and Child–Pugh B or C using the HS8_10 pulse as well as the HS8_5 pulse, with a high significance level (Kruskal–Wallis,Pb0.01).

Fig. 4.Scheme of the amplitude and frequency modulation function for HS8_10 and HS8_5 locking of 20 ms duration. HS8_10 locking consists of two segments a) and HS8–5 consists of four segments b), with an MLEV-type phase cycle.

Fig. 5.Simulated longitudinal magnetization after each spin lock pulse at TSL 20 ms: a) block pulse, b) HS8_10, and c) HS8_5. Relative B1 amplitude is on the vertical axis, andΔB0 is on the horizontal axis, in kHz units. Theﬁnal longitudinal magnetization after locking with HS8_10 and HS8_5 is shown to be more homogeneous for a wide range of relative B1 amplitude and ΔB0 compared to using block pulses. The contour lines in the plot are per 0.2 Meq interval, with dark blue equal to–Meq and yellow equal to +Meq. (For interpretation of the references to colour in thisﬁgure legend, the reader is referred to the web version of this article.)

(6)

4. Discussion

We investigated the effectiveness of applying a module consisting of stretched adiabatic pulses as a locking pulse to acquire robust T_1ρmaps at 3 T. Using the block pulse, which is standard in most spin locking

sequences for whole body systems, we found severe artifacts on the source images. The use of adiabatic pulses reduces B1 inhomogeneity- related artifacts, which cause fatal errors in calculating T_1ρvalues.

Witschey et al. reported a spin locking pulse that mitigates B0- and B1-type artifacts at allﬁeld strengths[26]. This type of B0 and B1 Fig. 6.Source images for each TSL and the calculated T_1ρmaps. The typical source images for each spin lock type are shown: block pulse (top), HS8_10 (middle), and HS8_5 (bottom) with the resulting T1ρmap (right). Using the block pulse, there are severe artifacts in segment 3 and segment 4 on each TSL image (white arrow) a), and severe artifacts in segment 7 and segment 8 on each TSL image (white arrow) b).

(7)

insensitive spin lock pulse (90(x)–τ/2(y)–180(y)–τ/2_(−y)–90(x)) was used for a neurological study. However, in our experience, it is not well suited for the liver due to the size of theﬁeld of view of the liver compared to the brain and the heart.

The diagram of thefinal longitudinal magnetization simulated using the Bloch equations after presenting each type of spin lock pulse to the equilibrium magnetization clearly shows that the block pulse is very sensitive to B0 and B1 inhomogeneity (Fig.5). This will cause severe artifacts on the image. In contrast, adiabatic spin locking results in a ho- mogeneousfinal magnetization for a wide range of B0 and B1fields. In particular, HS8_5 maintains homogeneity; therefore, it is useful for big organs, like the liver, where B0 and B1 are inhomogeneous, especially

at 3 T and higherfield strengths. However, there is no significant difference in visual scoring between HS8_10 and HS8_5. Thus, we can con- clude that the HS8_10 is working as well as the HS8_5 within the range of B0 and B1 variations expected for a liver examination on a clinical 3 T system like the one we used. The difference in visual scoring between the block and the two types of adiabatic spin locking pulse corroborates thefindings from the simulation. The stretched-type adiabatic spin locking pulse was able to provide homogeneous T_1ρimages.

Spin locking using adiabatic pulses represents a promising improvement to overcome B1 and B0 inhomogeneity, especially for body T_1ρ imaging.

Both mean T_1ρvalues obtained from the block and the two types of adiabatic spin locking pulse increased as liver function worsened. This result was similar to those previously reported[20,21,24]. However, the fundamental mechanisms leading to T_1ρprolongation in liverﬁbro- sis and cirrhosis have not been fully investigated[24]. Wang et al.[20]

described the usefulness of T_1ρrelaxation for the evaluation of liverfibrosis in a rat biliary duct ligation model. Takayama et al.[40]claimed that T_1ρrelaxation of the liver was not significantly correlated with liverfibrosis or with necro-inflammation. Thus, they concluded that T_1ρhas potential as a biomarker for the assessment of liver function, al- though it may not be suitable to estimate liverfibrosis or necro-inflammation. The present study did not investigate the mechanisms of T_1ρ prolongation with liver dysfunction. Liver dysfunction may contribute to T_1ρprolongation through combined mechanisms, such as biological, chemical, and physical factors, as Takayama et al. suggested.

The obtained T_1ρvalues using the adiabatic pulses were longer than those using the block pulse, similar to the results obtained by Jokivarsi et al.[41]in normal rat brain parenchyma, even if the average B_effis equal, as for the HS8_10 pulse. We conjecture that this difference in T_1ρis because the Beffof the adiabatic pulse is continuously changing during the pulse (Fig. 3), in contrast to the block pulse. The HS8_5 has a higher average Beff, therefore a different T_1ρcan be expected[32,42].

Furthermore, temporal and spatial variation of Beffwill result in a vary- ing T1ρrelaxation rate, which is difﬁcult to control or predict. One must be aware of this when interpreting the T_1ρvalues, and this should thus be done with extreme caution. Moreover, liver tissue contains a variety of macromolecules causing a mix of T1ρvalues within the voxels. Fur- ther, it is known that exchange and diffusion effects occur during Fig. 7.Results of visual scoring. The asterisk (*) represents a value ofPb0.001 between

block and HS8_5 and between block and HS8_10. There is no signiﬁcant difference between HS8_10 and HS8_5 (P= 0.792).

Table 2

Difference in mean T1ρvalues and range for normal tissue between each spin lock type.

Spin lock pulse Block HS8_10 HS8_5

Mean T1ρ(ms) Range of T1ρ(ms) Standard deviation (ms)

37.07⁎ 32.47–42.41

±3.30

46.12⁎ 44.03–48.77

±1.78

49.99⁎ 47.97–51.43

±2.14

⁎ Pb0.0001.

Fig. 8.Comparison of T1ρvalues between block, HS8_10, and HS8_5. The graphs indicate mean liver T1ρvalues and standard deviation of Child–Pugh A, Child–Pugh B or C, and normal using each spin lock pulse. Each spin lock type is shown on the horizontal axis; T1ρvalues are on the vertical axis.

(8)

locking and can affect theﬁnal signal. We did not study this aspect within the scope of this project.

It is thus difficult to achieve a meaningful absolute quantitative T_1ρin the clinical setting because the Beffvaries temporally and spatially. How- ever, using adiabatic locking pulses provides a locking that is effective over a largefield of view and thus enables the calculation of homogeneous T_1ρmaps. The visual scores of the T_1ρmaps acquired with adiabatic pulse locking are indeed higher compared to traditional block pulse locking. The visual scores for the respective Child–Pugh grades were not significantly different between HS8_10 and HS8_5. Further- more, the use of HS8_10 and HS8_5 pulses allows to differentiate between normal and Child–Pugh B or C with high statistical significance.

There is a limitation in the present study. The pulse duration of the adiabatic pulse was determined by the modulation function and opti- mal adiabatic condition. The total locking pulse is limited to an even number of adiabatic pulse segments (Fig. 4), and thus the TSL duration was restricted to an even multiple of that segment duration. In this study, HS8_10 uses 10 ms per adiabatic pulse segment, and the TSLs were set to 20 ms and 40 ms. However, we believe that this restriction to the TSLs was not disadvantageous for the calculation of the T_1ρ value of the liver, as mentioned in previous research[20,21,24].

In conclusion, we demonstrated that the stretched-type adiabatic spin locking method provides homogeneous and reduced artifact T_1ρ liver images and is useful for a robust evaluation of liver function using T_1ρ.

Disclosure of conﬂicts of interest

The authors have a conﬂict of interest to declare. Tomoyuki Okuaki and Marc Van Cauteren are Philips Healthcare employee. Tetsuo Ogino and Makoto Obara are Philips Electronics Japan employee. Yukihisa Takayama, Akihiro Nishie, Hiroshi Honda and Toshiaki Miyati have no relevant conﬂicts of interest to disclose.

References

[1]Child CG, Turcotte JG. Surgery and portal hypertension. Major Probl Clin Surg 1964;

1:1–85.

[2]Pugh R, Murray-Lyon I, Dawson J. Transection of the esophagus for bleeding esoph- ageal varices. Br J Surg 1973;60:649–54.

[3]Kamimura K, Fukukura Y, Yoneyama T, Takumi K, Tateyama A, Umanodan A, et al.

Quantitative evaluation of liver function with T1 relaxation time index on Gd- EOB-DTPA-enhanced MRI: comparison with signal intensity-based indices. J Magn Reson Imaging 2014;40(4):884–9.

[4]Katsube T, Okada M, Kumano S, Hori M, Imaoka I, Ishii K, et al. Estimation of liver function using T1 mapping on Gd-EOB-DTPA-enhanced magnetic resonance imaging. Invest Radiol 2011;46(4):277–83.

[5]Katsube T, Okada M, Kumano S, Imaoka I, Kagawa Y, Hori M, et al. Estimation of liver function using T2* mapping on gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid enhanced magnetic resonance imaging. Eur J Radiol 2012;81(7):

1460–4.

[6]Faria S, Ganesan K, Mwangi I. MR imaging of liverﬁbrosis: current state of the art 1.

Radiographics 2009;29:1615–35.

[7]Zhao F, Deng M, Yuan J, Teng G-J, Ahuja AT, Wang Y-XJ. Experimental evaluation of accelerated T1rho relaxation quantiﬁcation in human liver using limited spin-lock times. Korean J Radiol 2012;13(6):736–42.

[8]Sirlin CB. Science to practice: can T1rho imaging be used to diagnose and assess the severity of hepaticﬁbrosis? Radiology 2011;259(3):619–20.

[9]Boll D, Merkle E. Diffuse liver disease: strategies for hepatic CT and MR imaging 1.

Radiographics 2009;29:1591–614.

[10]Manduca A, Oliphant T. Magnetic resonance elastography: non-invasive mapping of tissue elasticity. Med Image Anal 2001;5:237–54.

[11]Yoshioka K, Hashimoto S. Can non-invasive assessment of liverﬁbrosis replace liver biopsy? Hepatol Res 2012;42:233–40.

[12] Yin M. Abdominal magnetic resonance elastography. J Acoust Soc Am. Acoustical So- ciety of Americ. 2012;132(3):1981.

[13]Nishioka H, Hirose J, Nakamura E, Okamoto N, Karasugi T, Taniwaki T, et al. Detecting ICRS grade 1 cartilage lesions in anterior cruciate ligament injury using T1ρand T2

mapping. Eur J Radiol 2013;82(9):1499–505.

[14]Hirose J, Nishioka H, Nakamura E, Oniki Y, Yamashita Y, Mizuta H. T1ρand T2map- ping of the proximal tibioﬁbular joint in relation to aging and cartilage degeneration.

Eur J Radiol 2012;81:2776–82.

[15]Nishioka H, Hirose J, Nakamura E, Oniki Y, Takada K, Yamashita Y, et al. T1ρand T2

mapping reveal the in vivo extracellular matrix of articular cartilage. J Magn Reson Imaging 2012;35(1):147–55.

[16]Takayama Y, Hatakenaka M, Tsushima H, Okazaki K, Yoshiura T, Yonezawa M, et al.

T1ρis superior to T2mapping for the evaluation of articular cartilage denaturaliza- tion with osteoarthritis: radiological-pathological correlation after total knee arthroplasty. Eur J Radiol 2013;82(4):e192–8.

[17]Mosher TJ, Zhang Z, Reddy R, Boudhar S, Milestone BN, Morrison WB, et al. Knee articular cartilage damage in osteoarthritis: analysis of MR image biomarker reproduc- ibility in ACRIN-PA 4001 multicenter trial. Radiology 2011;258(3):832–42.

[18]Wang Y-XJ, Zhao F, Yuan J, Mok GS, Ahuja AT, Grifﬁth JF. Accelerated T1rho relaxation quantiﬁcation in intervertebral disc using limited spin-lock times. Quant Imag- ing Med Surg 2013;3(1):54–8.

[19]Witschey WR. Zsido G a, Koomalsingh K, Kondo N, Minakawa M, Shuto T, et al. In vivo chronic myocardial infarction characterization by spin locked cardiovascular magnetic resonance J Cardiovasc Magn Reson 2012;14:37.

[20] Wang Y-X, Yuan J, Chu E, Go MYY, Huang H, Ahuja A, et al. T1ρMR Imaging Is Sen- sitive to Evaluate Liver Fibrosis : An Experimental Study in a Rat Biliary Duct Ligation Model 1 Purpose : Methods : Results : Radiology. 2011;259(3):712–9.

[21]Zhao F, Wang Y-XJ, Yuan J, Deng M, Wong HL, Chu ESH, et al. MR T1ρas an imaging biomarker for monitoring liver injury progression and regression: an experimental study in rats with carbon tetrachloride intoxication. Eur Radiol 2012;22(8):

1709–16.

[22] Yang Q, Yu T, Yun S, Zhang H, Chen X, Cheng Z, et al. Comparison of Multislice Breath-Hold and 3D Respiratory Triggered T1ρImaging of Liver in Healthy Volun- teers and Liver Cirrhosis Patients in 3 . 0 T MRI. 2016;(107):1–8.

[23]Hu G, Quan X, Lin X, Chan Q, Mei Y, Zhang X, et al. Assessment of liverﬁbrosis in rats by MRI with apparent diffusion coefﬁcient and T1 relaxation time in the rotating frame. J Magn Reson Imaging 2015;23:2015.

[24]Allkemper T, Sagmeister F, Cicinnati V, Beckebaum S, Kooijman H, Kanthak C, et al.

Evaluation ofﬁbrotic liver disease with whole-liver T1ρMR imaging: a feasibility study at 1.5 T. Radiology 2014;271(2):408–15.

[25]Rauscher I, Eiber M, Ganter C, Martirosian P, SaﬁW, Umgelter A, et al. Evaluation of T1ρas a potential MR biomarker for liver cirrhosis: comparison of healthy control subjects and patients with liver cirrhosis. Eur J Radiol 2014;83(6):900–4.

[26]Witschey WRT, Borthakur A, Elliott MA, Mellon E, Niyogi S, Wallman DJ, et al. Arti- facts in T1ρ-weighted imaging: compensation for B1 and B0ﬁeld imperfections. J Magn Reson 2007;186(1):75–85.

[27] Silver M., Joseph R., Hoult D. Highly selective andπpulse generation. J Magn Reson (1969). 1984. p. 347–51.

[28]Garwood M, DelaBarre L. The return of the frequency sweep: designing adiabatic pulses for contemporary NMR. J Magn Reson 2001;153(2):155–77.

[29]Uǧurbil K, Garwood M, Rath AR, Robin Bendall M. Amplitude- and frequency/phase- modulated refocusing pulses that induce plane rotations even in the presence of inhomogeneous B1ﬁelds. J Magn Reson 1988;78(3):472–97.

[30]Conolly S, Nishimura D, Macovski A. Sweep-diagram analysis of selective adiabatic pulses. J Magn Reson 1989;83(3):549–64.

[31]Uǧurbil K, Garwood M, Robin Bendall M. Amplitude- and frequency-modulated pulses to achieve 90° plane rotations with inhomogeneous B1ﬁelds. J Magn Reson 1987;72(1):177–85.

[32]Taheri S, Sood R. Spin-lock MRI with amplitude- and phase-modulated adiabatic waveforms: an MR simulation study. Magn Reson Imaging 2006 Jan;24(1):51–9.

[33]Michaeli S, Sorce DJ, Springer CS, Ugurbil K, Garwood M. T1ρMRI contrast in the human brain: modulation of the longitudinal rotating frame relaxation shutter- speed during an adiabatic RF pulse. J Magn Reson 2006;181(1):135–47.

[34] Mangia S, Liimatainen T, Garwood M, Michaeli S. Rotating frame relaxation during adiabatic pulses vs . conventional spin lock : simulations and experimental results at 4 T. Magn Reson Imaging 2009;27(8):1074–87.

[35] Casula V, Autio J, Nissi MJ, Auerbach EJ, Ellermann J, Lammentausta E, et al. Valida- tion and Optimization of Adiabatic T1ρand T2ρfor Quantitative Imaging of Articular Cartilage at 3 T. 2016;0:1–11.

[36]Deng M, Zhao F, Yuan J. Ahuja a. T, Wang YXJ. Liver T1ρMRI measurement in healthy human subjects at 3 T: a preliminary study with a two-dimensional fast-ﬁeld echo sequence. Br J Radiol 2012;85:1–6.

[37]Levitt MH, Freeman R, Frenkiel T. Broadband heteronuclear decoupling. J Magn Reson 1982;47(2):328–30.

[38]Yuan J, Zhao F, Chan Q, Wang Y-XJ. Observation of bi-exponential T1ρrelaxation of in-vivo rat muscles at 3T. Acta Radiol 2012;53:675–81.

[39]Kanda Y. Investigation of the freely available easy-to-use software“EZR”for medical statistics. Bone Marrow Transplant 2013;48(3):452–8.

[40]Takayama Y, Nishie A, Asayama Y, Ushijima Y, Okamoto D, Fujita N, et al. T1ρrelax- ation of the liver: a potential biomarker of liver function. J Magn Reson imaging. J Magn Reson Imaging 2015;42:188–95.

[41]Jokivarsi KT, Niskanen J, Michaeli S, Gröhn HI, Garwood M, Kauppinen RA, et al.

Quantitative assessment of water pools by T 1 rho and T 2 rho MRI in acute cerebral ischemia of the rat. J Cereb Blood Flow Metab 2009;29(1):206–16.

[42]Akella SVS, Regatte RR, Wheaton AJ, Borthakur A, Reddy R. Reduction of residual di- polar interaction in cartilage by spin-lock technique. Magn Reson Med 2004;52:

1103–9.